Chronic
Conditions,
Socioeconomic
Risks,
and
Behavioral
Problems
in Children
and
Adolescents
Steven
L. Gortmaker,
PhD;
Deborah
K. Walker,
EdD;
Michael Weitzman, MD; and Arthur M. Sobol, MA
From the Department of Behavioral Sciences, Harvard School of Public Health, and the
Boston University School of Medicine, Boston, Massachusetts
ABSTRACT. Children with a chronic health condition have long been considered at excess risk for psychosocial morbidity. Despite an increasing prevalence of chronic
childhood conditions and heightened concerns for the
quality of life of the chronically ill, population-based studies of behavior problems among children with chronic physical conditions are rare. Findings on the epidemiol-ogy of behavior problems in a nationally representative
sample of 11 699 children and adolescents aged 4 to 17
years in the United States are reported. Data included a
32-item parent-reported behavior problem index,
meas-ures of chronic childhood conditions, measures of school
placement and performance, and sociodemographic
van-ables. Analyses confirmed that chronic physical condi-tions were a significant risk factor for behavior problems,
independent of sociodemographic variables. Among
chil-dren these differences were observed across all subscales; among adolescents the largest differences were found for
the Depression/Anxiety and Peer Conflict/Social
With-drawal subscales. Rates of extreme behavior problem
scores (those in the top 10th percentile) were 1.55 times higher among children with a chronic health condition compared with children without a chronic condition (95% confidence interval 1.29 to 1.86). These independent odds were lowered to 1.44 when covaniates for confounding were introduced via a multivaniate logistic regression. Other independent risks included the absence of either biologic parent (odds ratio 2.05), male gender (1.53), low vs high family income (1.30), low vs high maternal edu-cation (1.51), and young vs old maternal age at childbirth
(2.57). Chronic health conditions were also a major risk factor for placement in special education classes and having to repeat grades. Despite evidence for effective interventions, health services for children with chronic conditions-particularly mental health services-remain fragmented, signaling the need for increased attention to behavioral problems and their treatment among all health professionals caring for children. Pediatrics
1990;85:267-Received for publication Feb 17, 1989; accepted Apr 25, 1989. Reprint requests to (S.L.G.) Dept of Behavioral Sciences, Har-yard School of Public Health, 677 Huntington Aye, Boston, MA 02115.
PEDIATRICS (ISSN 0031 4005). Copyright © 1990 by the
American Academy of Pediatrics.
276; chronic conditions, behavioral problems, socioeco-nomic status.
ABBREVIATIONS. BPI, Behavior Problem Index; NCHS, Na-tional Center for Health Statistics.
Children and adolescents with chronic health conditions have long been considered at substantial risk for excess psychosocial morbidity.’2 Although much of the evidence for excess risk comes from clinic-based studies,3 a number of population-based analyses have confirmed these relationships among children with chronic physical conditions in the Isle of Wight, England,’ two medium-sized urban areas in the United States,45 and Ontario, Canada.6
In contrast, some studies indicate little risk among clinic-based samples of chronically ill children.78
Concerns about the functioning of children with
chronic health conditions and their quality of life
have been heightened by recent increases in the
prevalence of chronic childhood conditions, in
many cases because of improved survival rates.9
The human immunodeficiency virus epidemic is now contributing substantially to the chronically
ill population of children, and further increases are
expected.
There are a variety of reasons to expect increased
psychosocial morbidity among these children. Many chronic health impairments are associated with chronic or recurrent episodes ofpain or acutely
diminished or altered physiologic function, which may promote anxiety, depression, or altered
the risk of problems with normal psychosocial ad-justment.2 In addition to their physical condition, children and adolescents with a chronic condition must cope with their own emotional reactions to
the illness and its care as well as to the reactions
of family members, friends, teachers, and others.’#{176}
The additional stress created for both parent and
child by the chronic condition and by the reactions of others is frequently manifested as emotional and behavioral problems or difficulties in social rela-tionships and self-esteem.
Factors such as poverty, family structure, and
parental characteristics can also exert important
influences upon behavior problems in children1’ via
a wide variety of etiologic mechanisms. For exam-pie, growing up within an environment of income
poverty could lead to symptoms of anxiety and depression by increasing exposure to stressful
ex-peniences and by decreasing psychological and
so-cial coping resources, in a manner similar to that
hypothesized for adults.12 Or a parent’s low level of
education could be associated with poor parenting
skills, which in turn could lead to behavior prob-lems. Divorce can produce stresses with negative
emotional consequences for children,’3 as can the
stress of a teenage pregnancy. Risks of social
mal-adaptation and psychologic well-being have also
been linked to family structure-in particular,
fe-male-headed single-parent families-among
chil-then with and without chronic conditions.14’5
In this analysis we examine the relationship be-tween parent-reported behavior problems of chil-then and adolescents and the presence of a chronic health condition. We control for the possible
con-founding of this relationship with the social and
economic status of the child’s family. The analyses address the following questions: (1) Are children
and adolescents with a chronic health condition at increased risk for behavior problems? Relative to children at social and economic disadvantage, are children with a chronic health condition at greater
risk? (2) Are children from poor, educationally
dis-advantaged, minority, or single-parent families who
also have a chronic health condition at particularly
increased risk? (3) For children at risk, what is
their pattern of utilization of mental health serv-ices?
METHODS
Data from the 1981 National Health Survey and Child Health Supplement were used in the analysis. The National Health Survey uses a complex,
mul-tistage probability sampling design to provide rep-resentative samples of the civilian
noninstitution-alized population of the United States. The 1981
Survey included a Child Health Supplement, which
collected data on 1 child in each eligible household, including 11 699 children and adolescents aged 4 to
17 years. The interview contained a series of
ques-tions concerning parent-reported behavior prob-lems and chronic childhood conditions,1#{176} as well as
sociodemographic measures. Thus, all information
is derived from parent reports; there were no actual
medical examinations of children or reviews of
med-ical records.
Definitions of Variables
Chronic Health Conditions. Data from a 59-item
chronic health condition checklist were used to
derive 19 categories of chronic conditions that were
considered to be validly reported by parents and that were potentially serious. We included only conditions that had been present for more than 3 months and had not been cured. We excluded con-ditions that were less likely to be serious, including allergies, tonsillitis, tuberculosis, pneumonia, acne, skin rashes, migraines, and headaches. Autism and mental retardation were excluded because of their strong relationship to psychosocial problems.8
Table 1 contains a list of the 19 chronic health
conditions studied and question wordings.
Studies have found that parent reports tend to overestimate the prevalence of clinically diagnosed chronic conditions; this overreporting declines with the severity or perceived stigma of the
condi-tion.17’8 Of greater concern in the present study is the possibility that underreporting of chronic con-ditions was more common among the poor because of lack of access to health services; this type of reporting bias would attenuate the relationship
be-tween social class variables and chronic conditions. Behavior Problems. The Behavior Problem Index (BPI), developed by Zill, was designed to encompass
domains of behavior similar to those covered by the
Achenbach Child Behavior Checklist,19 but to be much shorter. Most items used in the BPI were adapted from the parent-administered Child Be-havior Checklist and were chosen because of their reliability, high loading on the subscales of the Child Behavior Checklist, and their adaptability to an interview situation.2#{176} The Child Behavior
Checklist has been extensively validated and
ap-pears to discriminate moderately well between
chil-dren who are referred for clinical help with behavior
problems and children who are not so referred.’#{176} The BPI scales included an overall behavior problem score for each of two age groups (4 to 11 and 12 to 17), with six subscales for ages 4 to 11 and five subscales for ages 12 to 17. In computing
0.9 0.3 5.4 19.4
1.0 0.4
3.0
1.6
6.1
2.1
9.6
0.3 0.9 8.4
89.0
TABLE 1. Estimated Prevalence of Selected Chronic
Conditions Among Children and Adolescents Aged 4 to
17 Years: 1981 National Health Interview Survey*
Condition Categories (N) Estimated
Prevalence per 1000
Arthritis (40) 3.4
Asthma (342) 29.3
Blindness (39) 3.3
Cancer (any kind) (7) 0.6
Cardiac (8): rheumatic fever, 0.7
rheumatic or congenital heart disease
Cerebral palsy (11) Cystic fibrosis (4) Deafness (63)
Deformed body part (227): back, foot, leg, fingers, hand, arm, other deformi-ties
Diabetes (12)
Ear, nose, throat (5): cleft palate, harelip
Epilepsy (36): convulsions (repeated), seizures (re-peated)
Gastrointestinal (18): colitis, ulcer (excluding skin) Hearing (71): trouble
hear-ing/one ear, trouble hear-ing/both ears
Missing body part (24): fin-ger, hand, arm, toe, foot, leg
Orthopaedic (113): curvature of spine, clubfoot
Paralysis (4)
Sickle cell anemia (10) Vision (98): cataracts,
trou-ble seeing/one eye, trouble seeing/both eyes, other trouble seeing
Any of 19 chronic condition groups (1036)
* For each of the condition categories, subjects were
in-cluded only if the condition had been present more than
3 months and it had not been cured. Subjects who were
reported by parents to be autistic or mentally retarded, along with those with a wide variety of other conditions, were excluded.
were then summed to give total and subscale scores.
N. Zill (communication, 1985) has reported that the scale items were designed to represent measures of some of the “common syndromes of problem behavior found in children and adolescents.” The subscales included the following: Headstrong (eg, is disobedient at home); Antisocial Behavior (eg, bul-lies or is cruel or mean to others); Anxious/Dc-pressed Mood (eg, is unhappy, sad, or depressed); Hyperactive Behavior (eg, has difficulty concen-trating, cannot pay attention for long); Peer Con-flict/Social Withdrawal (eg, has trouble getting
along with other children); and (ages 4 to 11 only)
Immature Dependency (eg, cries too much).
Factor analyses of the items using principal
com-ponents followed by a principal components analy-sis with varimax rotation were carried out for a series of populations: separately for males and fe-males aged 4 to 11 and 12 to 17; separately for subjects with and without chronic conditions by sex for ages 4 to 11 and 12 to 17; and separately for subjects who had or had not ever seen a psycholo-gist, psychiatrist, or psychoanalyst, ages 4 to 1 1 and 12 to 17. The unrotated principal components
analysis within each age, gender, and chronic con-dition grouping confirmed the presence of a major first factor as reported by Zill (communication,
1985). This first factor explained 22% to 30% of the variance. All but a few of the items loaded on this first factor at a level of .40 or greater. The rotated factor analyses by subgroup confirmed the clustering of items identified by Zill, with the cx-ception of no clear-cut peer conflict/social
with-drawal grouping among children aged 4 to 11. A
confirmation was generally indicated by a loading of the items on a factor of .40 or higher.
Estimates of internal consistency reliability of the scales were made by using coefficient a.21 These
analyses indicated good reliability of the scales, for
children and adolescents both with and without a chronic health condition, particularly given the small number of items in the subscales (four to six).21 Reliability averaged .69 across the subscales for children and adolescents without a chronic con-dition and .70 for those with a condition. The total
BPI had an estimated reliability of .90, for subjects
both with and without a condition. Missing values on items making up the behavioral scales were accounted for by weighting observed scores. Miss-ing items were few in number-an average of 1.8% of responses were missing for any given item.
In addition to looking at mean differences in behavior problem scores, we also derived a
dichot-omous variable that identified the approximate top 10th percentile of subjects with extreme behavior problem scores; the cutoff was a score of 14 or more behavior problems (“sometimes” or “often” re-sponses) on the BPI. This cutoff score was substan-tially related to whether subjects had been referred
to mental health professionals (reported in the
Re-sults section), and this discrimination can be seen
as a validation of the BPI.’9
It is important to note that the BPI and other
scales such as the Child Behavior Checklist were designed to describe the behavior of children and
adolescents as seen by parents and were not meant to be used solely as the basis for diagnostic
infer-ences.1#{176}
were reported to have one of three levels of activ-ity limitation due to chronic conditions as defined by the National Center for Health Statistics (NCHS),16 depending on the extent to which their activities were limited as a result of chronic condi-tions: (1) unable to carry on major activity, which for preschool children was taking part in ordinary play with other children and for school-age children was the inability to go to school; (2) limitation in amount or kind of major activity, which for pre-school children was a limitation in the amount or
type of play with other children and for others it was a limitation to certain types of schools or a limitation in school attendance, including special schools, or restrictions on time spent in school; and
(3) no limitation in major activity, but otherwise limited; preschool children were not classified in this category; school-age children were so classified if they were limited in participation in athletics or other extracurricular activities. In analyses we grouped categories 1 and 2 together.
School Placement and Functioning. A single
ques-tion was used to determine whether a child or adolescent aged 5 to 17 was attending special
classes: “Does go to a special class or get special help in school because of a disability or health problem?” Another question asked whether the child or adolescent had ever “repeated any grades for any reasons.” Finally, parents were asked whether the child or adolescent had “ever been suspended, excluded or expelled from school.”#{176}
Sociodemographic Variables. Three racial cate-gories (white, black, other) were used in analyses.
We used 1980 family income and mother’s years of
completed education as our measures of social class. These measures were used rather than an occupa-tional measure because of their clear theoretical relationship to the development of behavior prob-lems in children. In addition, among adults educa-tional attainment and family income appear to be the most potent predictors of adult psychosocial dysfunction, such as distress.22 In cross-tabulations
and in the logistic regressions, income and maternal
age were represented by five categories and educa-tion by four categories. Other control variables included region of the United States and number
of older siblings.
Statistical
Analysis
All survey responses were weighted when
calcu-lating means, proportions, and logistic regressions using the weights provided by NCHS, which reflect the probability of selection, nonresponse, and poststratification adjustments.16 t Tests were used to test for differences in means and 2 tests for
differences in proportions. Multiple linear regres-sions were estimated when the dependent variable was continuous; the adjusted differences were
com-puted from these multivariate regressions. Logistic regressions were estimated when the dependent variable was dichotomous23 using the SAS
CAT-MOD procedure to produce maximum likelihood
estimates.24 The antilogs of the coefficient esti-mates can be interpreted as odds ratios associated with the predictor variable.
Estimates of statistical significance were made assuming simple random sampling. The actual sam-pling design was stratified, multistage, and clus-tered,16 and the assumption of simple random sam-pling in this case will result in overestimates of
statistical significance.25 Previous behavioral items used by NCHS in national surveys of a similar nature appear to have design effects greater than 1.0 (estimated by Gortmaker).26 Therefore, we will
discuss only coefficients with significance levels of .01 or less. Covariates believed to be contributors to design effects, including race and sociodemo-graphic variables, were included in the multivariate regressions. The sample sizes given in the different tables vary because of missing values. There were fewer observations in the multivariate logistic models because of limitations in the computer
pro-gram; accordingly, observations with incomplete
data were dropped. In the multiple linear
regres-sions, missing data were incorporated into the models. We reestimated logistic models to take into
account missing data and found no significant
change in results.
RESULTS
Prevalence of Chronic Conditions
The estimates of prevalence of the 19 chronic
conditions outlined in Table 1 are similar to esti-mates from other population surveys with compar-ably worded items.”5’9’2729 As reported in Table 2, there were modest differences in the prevalence of
chronic conditions according to socioeconomic
characteristics, including age, gender, family struc-ture, family income, and maternal education. These modest differences were expected and are similar
to those reported in earlier population studies.1’5’29 No significant differences with region of the coun-try or number of siblings was found.
There was a modest relationship between the NCHS measure of activity limitations due to chronic conditions and our indicator of chronic conditions. Of children and adolescents with 1 or more of the 19 chronic conditions studied, 23%
were indicated to be functionally limited to the
Chronic
Conditions
and Behavior
Problems
For children aged 4 to 11, the mean BPI score
was 6.9; among adolescents aged 12 to 17, the mean was 5.7. For both age groups, scores on the BPI and its subscales indicated more behavior problems among subjects with one or more chronic conditions
compared with those without a condition; the
un-adjusted mean differences are reported in Table 3.
10.0 .0001 Because the BPI is the sum of positive responses
7.7 to each item, mean differences can be interpreted
8 9 47 in terms of the mean number of endorsed items.
9:1 Among 4- to 11-year-old children, there was an
7.0 average of 1.7 more problems indicated among chil-dren with a chronic health condition. This excess
8.4 .04 occurred for all the subscales of the BPI. Among
9.4 adolescents the excess was on average 0.9
prob-8.2 .001 lems, and the largest differences were observed for 10.0 the Anxious/Depressed Mood and Peer Conflict/
Social Withdrawal subscales.
9.3 .10 Rates of extreme behavior problem scores were
1.55 times higher among children and adolescents
9:1 with a chronic condition compared with those
with-7.9 out a chronic condition (95% confidence interval
1.28 to 1.86) (see Table 4). Further analyses by 8.1 .03 individual condition (not shown) revealed higher
8 7 scores on the BPI (both means and extreme scores)
9:6 across virtually all the individual condition
cate-10.6 gories, supporting the notion of a noncategorical
approach,2’3’ although the small numbers in most
11.7 .05 categories limit the interpretation of estimates by
condition. There was an indication of substantially
8:9 higher rates of extreme BPI scores among children
8.5 and adolescents with epilepsy (35%; P < .0001; 95% confidence interval 20% to 51%).
.0001 45.5
49.0
7.2
TABLE 2. Prevalence of Any of 19 Chronic Conditions
by Sociodemographic Variables and Limitation of Usual
Activity Because of a Chronic Condition as Defined by
National Center for Health Statistics, 1981 National
Health Interview Survey, Ages 4 to 17 Years
Variables Prevalence (%) P Value
ofAny of 19 Chronic
Condi-tions
Gender Male (5968) Female (5731) Race
White (9661) Black (1729) Other (309) Age
4-11 (6332) 12-17 (5367)
Family structure
Both biologic parents (7582) Other (4117)
Mother’s age at childbirth 11-19 (1779)
20-24 (4191) 25-29 (3061) 30-34 (1555) 35+ (897) Mother’s education
Less than high school (3139)
High school diploma (5226) Some college (1759) Completed college (1342) Family income, 1980
$0-5000 (802) $5001-b 000 (1393) $10 001-15 000 (1521) $15 001-25 000 (2988) $25 000+ (4074)
Limitation of usual activity by chronic condition
Definitely cannot perform usual activity or can per-form but limited in amount and kind (208) Can perform usual activity
but limited in outside ac-tivities (283)
Not limited (11 208)
who were classified by NCHS as having some lim-itation due to a chronic condition, only 48% had 1 or more of the 19 chronic conditions. A conceptual
difficulty with the NCHS measure is that psycho-social problems are a major contributor to the
re-ported limitations. Newacheck et al.3#{176}have esti-mated that 10% of the NCHS activity-limiting chronic conditions are “mental or nervous system
disorders,” and approximately 15% are “impair-ments of speech, special sense or intelligence.33 In addition, our chronic condition measure included only a subset of the conditions used by NCHS.
Social Structure,
Chronic
Conditions,
and
Behavior
Problems
One competing explanation for higher rates of behavioral problems in children with chronic
con-ditions is that of disadvantaged socioeconomic
en-vironments: children with chronic conditions have been generally characterized as coming from poorer and less well-educated households than other chil-dren.32 These differences were confirmed in the
1981 NCHS sample, and the BPI was also
associ-ated with these variables, as indicated in Table 4. To estimate the independent association of be-havior problems with the presence of a chronic condition, controlling for these competing expla-nations, we estimated multiple linear and logistic regression equations as reported in Tables 3 and 5. The multivariate regressions lead us to reject the
hypothesis that these social structural variables
TABLE 3. Mean Behavior Problem Score Differences Between Children and Adolescents With and Without 1 or More of 19 Chronic Health Conditions, Unadjusted and Adjusted for Sociodemographic and Other Risks, by Subscales and Age, 1981 National Health Interview Survey
Ages and Subscales Unadjuste d Mean5 Adjusted Meant R2
Difference P Value Difference P Value
Ages 4-11 (N = 530 with
chronic health condi-tion, 5802 without)
BehaviorProblems 1.71 .0001 1.47 .0001 .06
Antisocial 0.24 .0001 0.18 .002 .07
Anxious/Depressed 0.37 .0001 0.28 .0001 .05
Headstrong 0.36 .0001 0.30 .0001 .04
Hyperactive 0.41 .0001 0.35 .0001 .06
Peer Conflict/Social With- 0.12 .0001 0.11 .0001 .02
drawal
Immature/Dependency 0.26 .0001 0.24 .0001 .03
Ages 12-17 (N = 506 with
chronic health condi-tion, 4861 without)
Behavior Problems 0.92 .0003 0.81 .001 .07
Antisocial 0.05 .36 0.04 .52 .07
Anxious/Depressed 0.37 .0001 0.30 .0001 .05
Headstrong 0.21 .004 0.18 .01 .04
Hyperactive 0.15 .02 0.13 .04 .06
Peer Conflict/Social With- 0.12 .0003 0.10 .002 .02
drawal
* Positive unadjusted differences indicate more behavior problems among subjects with a
chronic health condition; the actual number indicates the mean number of items with an affirmative answer.
t Differences adjusted via multiple linear regression including controls for age, race,
gender, family income, maternal education, maternal age, presence ofboth biologic parents, number of older siblings, and region.
controls were introduced, the mean differences in
the BPI associated with having a chronic condition
were similar to the uncontrolled differences (Table 3). The same results were found for all the
sub-scales. In all these multiple regression equations,
the R2 of the models is relatively low (.02 to .07). This fact reflects measurement error in the
behav-ioral measures as well as the limited predictive power of the variables included in the equations. Clearly, there are other unmeasured influences on behavior problems in addition to those identified in
this study.
Parallel results were observed for the logistic
regressions, which indicated that having a chronic
health condition was independently associated with an increased odds of behavioral problems of 1.44 (P < .0001) (see Table 5).
The estimated independent risk of behavioral
problems associated with a chronic condition was similar to the risk estimated for males (odds ratio 1.53), low vs high family income (odds ratio 1.30),
low vs high maternal education (1.51), and less than
that estimated for a young vs older mother (odds ratio 2.57) and for a family missing either biologic
parent (odds ratio 2.05).
Evidence
for Interactions
We tested for the presence of interactions if both a chronic condition and social disadvantage were
present by adding terms to the multivariate regres-sions indicating the combination of chronic condi-tions and low income (<$5000), chronic conditions
and low education (less than high school), chronic
conditions and the absence of one or both biologic parents, chronic conditions and being black, and chronic conditions and being male. We also tested
for interactions by including terms signifying
dif-ferent slopes of the income or mother’s education variables, depending on the presence of a chronic condition. Only the interaction of having a chronic
condition with the absence of either biologic parent
was marginally statistically significant in the
equa-tion predicting the BPI; this interaction magnified
the effect of a chronic condition if both biologic
parents were not in the household. This interaction
was not significant in the logistic regression pre-dicting extreme scores.
However, it should be recognized that all these
models indicate substantially greater risk for
TABLE 4. Prevalence of Extreme Scores (>14) on the Behavior Problem Index by Sociodemographic Variables, Presence of 1 or More of 19 Chronic Conditions, and Limitation of Usual Activity Because of a Chronic Condition as Defined by National Center for Health Statistics, 1981 National Health Interview Survey, Ages 4 to 17 Years
Variables % With Behavior P Value
Problem Scores >14
10.5 13.0 8.2
7.7 16.4
.0001
16.8 15.2 12.8 9.4 8.6
26.9 .0001
23.4
10.2 Gender
Male (5968) Female (5731) Race
White (9661) Black (1729) Other (309) Family structure
Both biologic parents (7582) Other (4117)
Mother’s age at childbirth 11-19 (1779)
20-24 (4071) 25-29 (3061) 30-34 (1555) 35+ (897) Mother’s education
Less than high school (3139) High school diploma (5226) Some college (1759) Completed college (1342) Family income, 1980
$0-5000 (802) $5001-b 000 (1393) $10 001-15 000 (1521) $15 001-25 000 (2988) $25 000+ (4074)
Any of 19 chronic conditions Yes (1067)
No (10 975)
Limitation of usual activity by chronic condition
Definitely cannot perform usual ac-tivity or can perform but limited in amount and kind (208)
Can perform usual activity but
lim-ited in outside activities (283) Not limited (11 208)
12.7 .0001
8.7
15.4 12.2 8.6 7.9 5.0
.003
.0001
12.7 .0001
10.9 9.6 7.6
.0001
15.1 .0001
10.4
models, for example, are multiplicative in the odds. Therefore, the odds ratios in Table 5 indicate that a male adolescent with a chronic health condition
in a single-mother-headed household who gave
birth to her child before age 20 experiences up to 12 times the risk of behavior problems compared with other children.
Once socioeconomic differences were controlled, the relationship ofbeing black to behavior problems was reduced to statistical insignificance. (See Table
5; the linear regression results were similar.) These results persisted when interactions between race
and socioeconomic variables were added to the equation.
School
Placement
and Performance
and Chronic
Conditions
We also examined the association of chronic con-ditions with reports of school placements, having to repeat a grade, and being expelled or suspended. We estimated logistic regressions as before and controlled for sociodemographic variables. As
ex-pected, those with a chronic condition had
substan-tially higher relative odds of special education placement (odds ratio 2.65; P < .0001) and higher
rates of having to repeat a grade (odds ratio 1.38; P
< .0007). There was no independent association of
Top 10th Percentile of BPI Low BPIScores
- All Children
Male Female
4o
g
:
P40Chre,c Condition .27
.0001
.34
.0001
.0001
.01
.02
.-.-.---on 1.08
1.00
1.53 1.00
1.11 0.97 1.00
1.00 2.05
2.57 2.38 1.91 1.66 1.00
1.51 1.44 1.24 1.00
1.30 1.26 1.18 0.93 1.00
Presence of chronic Health Condition
Figure. Children and adolescents aged 4 to 17, with and without chronic conditions, reported by their parents ever to have seen a psychiatrist, psychologist, or psychoana-lyst. BPI, Behavior Problem Index.
TABLE 5. Estimated Odds Ratios From Logistic Regression Predicting Top 10th Percentile Scores (>14)
on the Behavior Problem Index, 1981 National Health
Interview Survey, Ages 4 to 17 Years (N = 10 325)
Odds Ratiot P Value
Predictive Variables and Categories*
Age 4-11 12-17 Gender
Race White Black Other Family structure
Both biologic parents Other
Mother’s age at childbirth 11-19
20-24 25-29 30-34 35+
Mother’s education Less than high school High school diploma Some college Completed college Family income, 1980
$0-5000 $5001-b 000 $10 001-15 000 $15 001-25 000 $25 000+
Any of 19 chronic condi-tions
Yes No
1.44 .0003
1.00
* Other variables that did not independently predict
in-cluded region of the country and number of siblings.
t Odds ratios independently associated with being in the top 10th percentile of the Behavior Problem Index, rela-tive to the reference category (which has a value of 1.00).
Visits to Psychologists
and Psychiatrists
We found a substantial correlation between
scores on the BPI and parental reports of subjects’
ever having seen a psychiatrist, psychologist, or
psychoanalyst about any emotional, mental, or
be-havioral problem (r = .28). Overall, 6% of children and adolescents aged 4 to 17 were reported by their
parents to have seen a psychiatrist, psychologist, or
psychoanalyst (see Figure). For children who scored
in the top 10th percentile of the BPI, the rate was
25% vs 4% for all other children (P < .0001). This difference of 21% was changed little by controlling for the variables outlined in Table 3 in a multiple regression analysis.
Among children with a chronic condition, 11% had visited a psychologist, psychoanalyst, or
psy-chiatrist. Of all children with a chronic condition
and an extreme score on the BPI, 38% had visited a mental health professional. While these data in-dicate that children at highest risk are receiving more mental health services, most children at risk (either by virtue of an elevated BPI score or because of a chronic condition) had not visited a psycholo-gist, psychiatrist, or psychoanalyst.
DISCUSSION
This study confirms the presence of a chronic health condition as a significant risk factor for behavioral problems in children and adolescents in the United States. This increased risk was inde-pendent of the socioeconomic, demographic, and
racial characteristics of the household.
In the aggregate, the 55% increase in risk found
is more modest than suggested by many earlier studies.6 Differences with clinic-based studies
could be expected for a number of reasons: clinics may overrepresent children and adolescents with
the most serious conditions; conditions involving brain abnormality-a major correlate of behavioral problems-may be overrepresented;8 families who use medical-center-based tertiary facilities may be poorer or have other risk factors for psychosocial function; or the classification in this study of a chronic condition may be weighted toward milder conditions. Another interpretation of this lessened relationship could focus on the effectiveness of recent public policies in the United States aimed at improving the social and educational environments of children with chronic conditions. The implemen-tation of PL 94-142, for example, while clearly far from ideal, has had the effect of improving the school-based experiences of many children in the
United States.33
Our estimates indicate that the risks of
poverty, low maternal education, and male gender.
We found no evidence that the risks of children and adolescents with a chronic condition who were
also poor or whose mother had little education were greater than the risks predicted based on these two independent factors alone.
Although low family income, low maternal edu-cation, and living with fewer than two biologic parents were risk factors for behavior problems, no association was found between being black and having behavior problems once socioeconomic dif-ferences were controlled. This finding is consistent with studies concerrning the mental health status of adults.12 These results suggest that racial
differ-ences in behavior problems are due to socioeco-nomic differences between the races.
This study is limited in a number of respects by incomplete data: we do not have a measure of maternal mental health, and previous studies have documented strong relationships between maternal depression and reports of behavior problems in
children. This relationship could provide another interpretation for the strong independent associa-tion found between family income and behavioral
problems in children and adolescents: poverty and associated stress has been implicated as an impor-tant causal variable in the generation of anxiety
and depression in women,22 and this fact could in
turn be responsible for elevated levels of behavioral problems in children, due either to the impact of maternal distress on children or to distorted
re-porting. The R2 statistics ofthe multivariate models confirm that much of the variance in these meas-ures of behavior problems remains unexplained.
We have already noted that the NCHS measures
of activity limitation were confounded with behav-ioral problems. However, the recent study by Cad-man et al6 in Ontario specifically excluded behavior problems as functionally limiting conditions (M.D., verbal communication, October 1987) and found
that functional impairment due to a chronic health condition was an important additional risk factor for psychosocial maladjustment.
Finally, our discussion of the interactive
influ-ences of socioeconomic disadvantage and chronic conditions is limited by the unavailability of other
family variables such as paternal criminality. Rutter34 has indicated how the confluence of six such familial factors may interact and has suggested
how these factors may further interact with
mdi-vidual differences.
IMPLICATIONS FOR POLICY AND PRACTICE
Our analysis did not find strong relationships between visits to psychologists and psychiatrists and either the parental income, education, or
child-hood chronic condition variables. Most children and adolescents at high risk for behavioral prob-lems had never seen a mental health professional, and 62% ofchildren with behavioral problem scores in the top 10th percentile and a chronic health condition had not done so. We do not have further data concerning these mental health visits, and we expect there to be substantial differences in site of treatment by social class, with poorer children re-ceiving most services within school-based
set-tings.35 In addition, some social workers, nurse cli-.
nicians, and pediatricians also provide mental health services, but we have no data on this prac-tice. Not surprisingly, many studies have found little use of mental health services,6’36 and one recent study found low use among children in spe-cial education placements who were labeled
emo-tionally disturbed.37
Although it has been generally accepted for some time that children with chronic conditions are at
higher risk for developing behavioral problems, the literature contains few accounts of systematic, well-documented attempts to prevent or ameliorate
these functional problems. There have been a
num-ber of studies of very small samples documenting
the effectiveness of educational programs geared to
increase health knowledge and self-care and to im-prove the functioning of children with asthma.38’39 In addition, there are a few programs that include
children with a broad range of chronic disorders. These include PACTS, a randomized controlled
trial of a university hospital-based home care pro-gram for children with long-term health needs, which demonstrated beneficial effects on psychol-ogic functioning of the children.4#{176} A lay family counselor intervention also improved psychosocial
functioning of children in subspecialty settings at
the University of Rochester.4’ A randomized con-trolled trial of a social work intervention conducted
at Montreal Children’s Hospital indicated no ef-fects of a 6-month intervention.42 A recent larger-scale program is the Rural Efforts to Assist Chil-dren at Home (REACH) in northern Florida. This
project has demonstrated the feasibility of imple-menting a rural outreach intervention in one state, but a rigorous evaluation was not designed along
with the program.
In summary, there is evidence that both targeted and comprehensive service interventions can im-prove the psychosocial functioning of children and adolescents with a chronic condition. The
contin-uing evidence for risk, the demonstrated need for
better coordinated and financed services,43 and the
success of some intervention studies to date signals
the need for further interventions and their careful evaluation. The patterns of use of mental health
for increased skills and concern for behavioral prob-lems among all health professionals caring for chil-dren and adolescents.
ACKNOWLEDGMENTS
This work was supported by grant 85103185 from the William T. Grant Foundation.
We thank members of the Research Consortium on
Chronic Illness in Childhood and anonymous reviewers for helpful comments.
REFERENCES
1. Rutter M, Tizard J, Wightmore K. Education, Health and
Behavior. Longon, England: Longmans; 1970
2. Pless IB, Pinkerton P. Chronic Childhood Disorder:
Pro-moting Patterns of Adjustment. London, England: HO
Klimpton Publishers; 1975
3. Nolan T, Pless TB. Emotional correlates and consequences of birth defects. J Pediatr. 1986;109:210-216
4. Pless IB, Satterwhite BB. Chronic illness. In: Haggerty RJ,
Roghmann KJ, Pless IB, eds. Child health and the
Corn-rnunity. New York, NY: John Wiley & Sons; 1975:78-94 5. Walker DK, Gortmaker SL, Weitzman M. Chronic Illness
and Psychosocial Problems Arnong Children in Genesee
County. Boston, MA: Harvard School of Public Health; 1981
6. Cadman D, Boyle M, Szatmari P, et al. Chronic illness, disability, and mental and social well-being: findings of the Ontario Child Health Study. Pediatrics. 1987;79:805-813 7. Drotar D, Doershuk CF, Stern RC, et al. Psychosocial
func-tioning of children with cystic fibrosis. Pediatrics. 1981; 67:338-343
8. Breslau N. Psychiatric disorder in children with physical
disabilities. J Arn Acad Child Psychiatry. 1985;24:87-94 9. Gortmaker SL, Sappenfield W. Chronic childhood disorders:
prevalence and impact. Pediatr Clin North Arn. 1984;31:3-18
10. Perrin EC, Ramsey BK, Sandler HM. Competent kids:
children and adolescents with a chronic illness. Child Care
Health Dev. 1987;13:13-32
11. Clarizo HF, McCoy GF. Behavior Disorders in School-Aged
Children. Scranton, PA: Chandler; 1970
12. Kessler RC, Cleary PD. Social class and psychological dis-tress. Arn Sociol Rev. 1980;45:463-478
13. Furstenberg FF Jr, Seltzer JA. Divorce and child develop-ment. Sociol Stud Child Dev. 1986;1:137-160
14. Stein REK, Jessop DJ. Relationship between health status and psychologic adjustment among children with chronic conditions. Pediatrics. 1984;73:169-174
15. Kellam SG, Ensminger, Turner J. Family structure and the mental health of children. Arch Gen Psychiatry. 1977;
34:1012-1022
16. National Center for Health Statistics. Current estimates from the National Health Interview Survey: United States 1981. Vital and Health Statistics. Series 10, No. 141. Wash-ington, DC: US Government Printing Office; October 1982. (PHS) 83-1569
17. Haggerty RI, Roghmann KJ, Pless lB. Child Health and the
Cornrnunity. New York, NY: John Wiley & Sons; 1975
18. National Center for Health Statistics. Net differences in interview data on chronic conditions and information de-rived from medical records. Vital and Health Statistics.
Series 2, No 57. Washington, DC: US Government Printing Office; 1973
19. Achenbach TM, Edelbrock C. Manualfor the Child Behavior
Checklist and Revised Child Behavior Profile. Burlington,
VT: Queen City Printers; 1983
20. Peterson JL, Zill N. Marital disruption, parent-child rela-tionwships, and behavior problems in children. J Marriage
Farnily. 1986;48:295-307
21. Nunnally J. Psychometric Theory. New York, NY: McGraw-Hill International Book Co; 1978
22. Kessler RC. A disaggregation of the relationship between socioeconomic status and psychological distress. Arn Sociol
Rev. 1982;47:752-764
23. Anderson S, Aquier A, Hanick W, et al. Statistical Methods for Comparative Studies. New York, NY: John Wiley & Sons; 1980
24. SAS Institute Inc. SAS/STAT Guide for Personal
Corn-puters. Version 6 ed. Cary, NC: SAS Institute mc; 1987
25. Landis JR, Lepkowski JM, Eklund SA, et al. A Statistical
Methodology for Analyzing Data From a Complex Survey:
The First National Health and Nutrition Examination
Sur-vey. Series 2, No 92. 1982. US Dept of Health and Human
Services publication (PHS) 82-1366
26. United States Department of Health, Education, and Wel-fare. Parent Ratings of Behavioral Patterns of Children:
United States. Rockville, MD; November 1971. US Dept of
Health, Education, and Welfare publication (HSM) 72-1010
27. Pless IB, Roghmann K. Chronic illness and its
conse-quences; observations based on three epidemiological sur-veys. J Pediatr. 1971;79:351-359
28. Walker DK, Gortmaker SL. Final Report: Community Child
Health Studies. Grant MC-R-250437. Springfield, VA: Na-tional Technical Information Service; 1983
29. Cadman D, Boyle MH, Offord DR, et al. Chronic illness and
functional limitation in Ontario children: findings of the
Ontario Child Health Study. Can Med Assoc J. 1986;
135:761-767
30. Newacheck PW, Budetti PP, Halfon N. Trends in activity-limiting chronic conditions among children. Am J Public Health. 1986;76:178-184
31. Stein R, Jessop D. A non-categorical approach to chronic childhood illness. Public Health Rep. 1982;97:354-362 32. Kovar MG. Health status of US children and use of medical
care. Public Health Rep. 1982;97:3-15
33. Singer JD, Butler JA. The Education for All Handicapped Act: schools as agents of social reform. Harvard Educ Rev.
1987;57:125-152
34. Rutter M. Invulnerability, or why some children are not damaged by stress. In: Shamsie SJ, ed. New Directions in
Children ‘s Mental Health. New York, NY: SP Medical & Scientific Books; 1979:53-75
35. Walker DK, Palfrey JS, Butler JA, et al. Health and com-munity services for children with impaired mobility: who supports what? Public Health Rep. In press
36. Staruch K, Breslau N, Weitzman M, et al. Use of health services by chronically ill and disabled children. Med Care.
1984;22:310-328
37. Walker DK. Educational options. In: Stein R, ed. New
Directions in Care of Children With Chronic Illness. New York, NY: Springer; 1988
38. Fireman P, Friday GA, Gira C, et al. Teaching self-manage-ment skills to asthmatic children and their parents in an ambulatory care setting. Pediatrics. 1981;68:341-348
39. Lask B, Matthew D. Childhood asthma: a controlled trial of family psychotherapy. Arch Dis Child. 1979;54:116-119 40. Stein REK, Jessop DL. Does pediatric home care make a
difference for children with chronic illness: findings from the pediatric ambulatory care treatment study. Pediatrics. 1984;73:845-853
41. Pless IB, Satterwhite B. Chronic illness in childhood: selec-tion, activities and evaluation of non-professional family counselors. Clin Pediatr. 1972;11:403-410
42. Nolan N, Zvagulis I, Pless B. Controlled trial of social work in childhood chronic illness. Lancet. 1987;411-415
43. Hobbs N, Perrin JM, Ireys HT. Chronically-Ill Children and
